System Dynamics Methodology
System dynamics is a computer-aided approach to policy analysis and design that applies to problems arising in complex social, managerial, economic, or ecological systems. The approach is appropriate for any dynamic system characterized by interdependence, mutual interaction, information feedback, and circular causality. System dynamics models are built around a particular problem. The problem defines the factors (i.e., relevant variables) that are to be included in the model. This represents the model’s boundary, which may cross departmental or organizational boundaries. One of the unique advantages of using system dynamics models to study public policy issues or problems is that they can easily be extended or revised to address additional questions as they arise.
System dynamics models rely on three sources of information: numerical data, the written database (reports, operations manuals, etc), and the expert knowledge of key participants in the system. The numerical database is very small, the written database is larger, and the expert knowledge of key participants is vast. System dynamicists rely on all three sources, with particular attention paid to the expert knowledge of key participants. Through the use of available data and verbal descriptions provided by experts, the modeling process exposes new concepts and/or previously unknown but significant variables.
System dynamics models are excellent tools to study problems that arise in closed-loop systems, systems in which conditions are converted into information that can be observed and acted upon in order to change the initial condition. For example, when the backlog of pending cases for district attorny's increases beyond a certain level it sends a signal to district attorny's that additional plea bargains may be required in order to reduce the backlog to an acceptable level. This completes a feedback loop. However, this same feedback loop could also work in the opposite direction. If the backlog of cases is below some desired or acceptable level it will indicate to district attorny's that the need to plea bargain, for purposes of reducing the backlog of cases, is no longer necessary and the willingness of district attorny's to offer plea bargains will be reduced.
Using a SD Model to Develop a Theory
A system dynamics model represents a theory about a particular problem. Since any model in the social sciences is only a theory, the most that can be attained from these models is that they be useful. System dynamics models are useful because the mathematical underpinning needed for computer simulation requires that the theory be precise. The process of combining numerical data, written data, and the knowledge of experts in mathematical form often identifies inconsistencies about how we think the system works. The model educates us by identifying these inconsistencies. Simulation allows us to see how the complex interactions we have identified in the model work when they are all active at the same time. This is what occurs in the real system. Furthermore, we can test a variety of policies quickly to see how they play out in the long run. The bottom line is that the model will represent a theory about what is causing the problem and what can be done to solve the problem.
Output of SD Models: Behavior versus Forecasts
People who take a systems view of policy problems know that the behavior generated by complex organizations cannot be well understood by examining the parts. By taking this holistic view, models capture time delays, amplification, and information distortion, as they exist in organizations. By developing computer simulation models that incorporate information feedback, systems modelers seek to understand the internal policies and decisions and the external phenomena that combine to generate the problems observed. They seek to predict dynamic implications of policy, not forecast the values of quantities at a given time in the future.
As such, system dynamics models do not provide numerical forecasts. Rather, system dynamics models are policy tools that examine the behavior of key variables over time. Historical data and performance goals provide baselines for determining whether a particular policy generates behavior of key variables that is better or worse, when compared to the baseline or other policies. Furthermore, models provide an explanation for why specific outcomes are achieved. Simulation allows us to compress time so that many different policies can be tested, the outcomes explained, and the causes that generate a specific outcome can be examined by knowledgeable people working in the system before policies are actually implemented.